[GitHub] CyberZHG edited a comment on issue #13557: Fix BatchNorm converter for CoreML when fix_gamma=True

[GitBox <gi...@apache.org>]

CyberZHG edited a comment on issue #13557: Fix BatchNorm converter for CoreML when fix_gamma=True
URL: https://github.com/apache/incubator-mxnet/pull/13557#issuecomment-451386266
 
 
   @Roshrini 
   
   The results are always exactly the same after setting the seed values.
   
   I triggered the CI (uncomment the line in `.travis.yml` file) for the converter and it can pass the tests with the original seeds and the new seeds. Therefore it is __not necessary__ to change the seed values to pass the CI tasks.
   
   Besides, I've done some calculations manually for convolution and batch normalization, and the results are the same as mxnet's outputs. However, there are always some slight differences between coreml's outputs and mxnet's outputs in my machine. I think there may be some environmental differences and the new seeds are more compatible with different versions.
   
   In summary, currently all tests could pass the CI tasks and I can change back the seed values if it is insisted.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services